Author: Chainlink Economics
In December 2024, Chainlink introduced Smart Value Recapture (SVR)—a novel oracle solution designed to enable DeFi applications to recapture the Maximal Extractable Value (MEV) derived from their use of Chainlink Price Feeds. The initial version of Chainlink SVR was built in collaboration with BGD Labs, Flashbots, and other contributors to the Aave DAO and will initially focus on enabling DeFi lending protocols to recapture oracle-related MEV from liquidations. MEV derived from the use of oracles is commonly referred to as Oracle Extractable Value (OEV), and this paper focuses only on backrunning transactions to capture liquidation OEV.
Recapturing the OEV generated by liquidations in lending protocols presents a potentially significant revenue opportunity for protocols such as Aave, where the right to liquidate undercollateralized positions is auctioned off and the value generated is returned back to the protocols and the oracles that generated the value. Notably, liquidations most commonly occur during times of high market volatility, with just a few liquidations compromising the vast majority of OEV.
To support the Aave and broader DeFi community, this report presents a framework for helping protocols parameterize their usage of Chainlink SVR, particularly related to how they can think about delaying oracle updates during the SVR auction process. We outline strategies protocols can consider to mitigate risks, such as adjusting liquidation bonuses and loan-to-value (LTV) ratios to help ensure protocol stability. The results suggest that, with proper calibration from integrating protocols, lending protocols like Aave v3 can benefit from integrating Chainlink SVR while still preserving their current risk profile.
Oracle Extractable Value (OEV) Calculation
To explore the risk dynamics of delayed liquidations, we collected liquidation events from January 1, 2024 to December 3, 2024 for AAVE v2 and v3 in loans involving the following assets: WETH, WBTC, and AAVE. Let L be the set of all the liquidations in our dataset, then for each liquidation i∈L, the initial OEV is calculated as:
Max OEVi=max(0,Collateral USDi–Debt USDi)Max OEVi=max(0,Collateral USDi–Debt USDi)
This represents the potential profit that can be extracted due to the difference between the amount of collateral and debt, as illustrated below. However, this calculation greatly overestimates the realized OEV, by assuming a perfectly efficient MEV market, while ignoring execution costs and risk.

Adjusted OEV
To adjust our estimates, we discount each liquidation based on the value it produces. The parameters α, β, and P play a critical role in refining the calculation of adjusted OEV. These variables provide a conservative estimate of the OEV that can be captured by Aave under different conditions.
α represents the capture rate applied to OEV values below the percentile threshold P, reflecting more common liquidation events where the competition is moderate and execution risks are lower. Β represents the capture rate applied to OEV values exceeding the percentile threshold P, which typically correspond to exceptional or large liquidation events that may involve greater execution costs or inefficiencies. α and β have been set to 0.5 and 0.3 respectively, representing a conservative yet realistic value based on real-life analysis performed by Chainlink Labs. Finally, P (Percentile Threshold) sets the boundary between ”common” and ”exceptional” OEV events. For this study, the 90th percentile is used, meaning the top 10% of OEV values are subject to the more conservative capture rate β to account for the increased complexity and market dynamics involved. These parameters ensure that the adjusted OEV estimates are not overly optimistic, allowing Aave to base its decision-making on conservative yet actionable projections.
Specifically:
OEVadji={α×Max OEVi,if OEVi≤Pβ×Max OEVi,if OEVi>POEViadj={α×Max OEVi,if OEVi≤Pβ×Max OEVi,if OEVi>P
Where:
- α and β act as a conservative proxy of the realized take rates (e.g., α = 0.5, β = 0.3).
- P is the percentile threshold, after which the take rate decreases. We choose the 90th percentile of OEV to err on the side of caution. The total adjusted OEV for all loans is:
Total OEVadj=∑i∈LOEVadjiTotal OEVadj=∑i∈LOEViadj
As the figure below demonstrates, the top 10% of liquidations represent roughly 80-90% of the liquidation volume. Therefore, our assumptions imply that 80-90% of the liquidation volume will be discounted at the maximum rate (i.e., 30% take rate).

Potential Risk Exposure Calculation
Oracle Delay Probability
The probability of a delay of d blocks is modeled as a geometric distribution:
Pdelay(d)=(1–f)d×fPdelay(d)=(1–f)d×f
The cumulative probability that the oracle is delayed by at least d blocks is:
Pdelay(≥d)=(1–f)dPdelay(≥d)=(1–f)d
This is based on f, which is the share of validators connected to the Flashbots infrastructure—currently estimated at 90% and has been generally constant over the last year. This implies that there is a 0.1% probability of a transaction not being included in the first three blocks. The proportion of validators utilizing Flashbots’ MEV solution is a critical factor influencing the time it takes to auction a transaction compared to submitting it directly to the public mempool.
Notably, if the fraction of validators using Flashbots declines, the probability of oracle updates being delayed increases. This could lead to longer windows during which liquidations are not executed, increasing risk exposure to lending protocols in volatile markets.
To account for this risk, we modeled delay probabilities based on the current fraction of validators connected to Flashbots, which is conservatively estimated at 90%. This modeling helps ensure that the risk from reduced validator participation is explicitly quantified. We also considered the worst-case price movements during delays and calculated their impact on bad debt. This approach ensures that even extreme outcomes stemming from external dependencies are incorporated into the framework.
Value at Risk (VaR) Thresholds
The analysis in this section aims to quantify certain risks associated with delayed oracle updates caused by the auction process, particularly focusing on their impact on the protocol’s exposure to bad debt. By using historical price data, we model the most extreme positive and negative price movements for each token during delays, capturing potential worst-case scenarios.
This approach is structured to evaluate the sensitivity of different tokens to delays and assess how these price movements might affect the liquidation process. Specifically, we determine the probability and magnitude of bad debt arising from delays, providing actionable insights into how lending protocols like Aave can adjust its parameters—such as liquidation bonuses—to offset these risks. The subsequent VaR calculations will highlight these risks across various tokens and delay durations, establishing a clear foundation for tailoring mitigation strategies to asset-specific characteristics.
Specifically, for each block delay d, the corresponding VaR threshold is determined based on historical price movements from January 1, 2022 to Dec 3, 2024, assuming a constant block time of 12 seconds:
VARd=percentile-based price movement over d blocks or Td=d×12 seconds.VARd=percentile-based price movement over d blocks or Td=d×12 seconds.
For each asset we identify the most extreme positive and negative price movements for each oracle delay (top and bottom 0.1% of the distribution).

This table highlights notable differences in price sensitivity across tokens during oracle delays, providing insights into how various market dynamics affect protocol risk. Tokens like WBTC and WETH exhibit relatively modest price changes over increasing delays, reflecting their larger market capitalizations and higher liquidity. These factors generally insulate such assets from sharp price swings, even during transient disruptions. In contrast, AAVE shows significantly higher VaR values, which can be attributed to the smaller market sizes and lower liquidity. These characteristics make such tokens more vulnerable to volatility, where even minor market activities can cause substantial price fluctuations.
Another key observation is the varying rate at which VaR values increase with longer delays. For instance, while WBTC experiences a gradual rise in VaR, indicating limited sensitivity to short-term disruptions, AAVE’s VaR escalates sharply with each additional delay. This emphasizes AAVE’s heightened exposure to market instability during periods of latency, underscoring the need for more proactive risk mitigation measures for such assets.
These differences in VaR across tokens underscore the importance of protocols tailoring risk management strategies to the specific characteristics of each asset. Stable tokens allow for more conservative adjustments to liquidation bonuses, while more volatile tokens require higher margins of safety to protect against potential bad debt. Understanding these nuances enables asset-specific measures that balance risk mitigation with the opportunity to capture oracle-related MEV during liquidations.
Per-Liquidation Bad Debt Calculation
Situation 1: Price decrease during oracle delay leads to collateral decreasing in value:
To analyze the adverse effects of a price decrease during an oracle delay, we simulate the most extreme negative price changes of collateral assets and explore their impact on historical liquidations, while keeping the value of the debt constant. Specifically, we calculate the depreciated collateral value in USD as follows:
Adjusted Collateral in USDi,d=(1–VARdecrease,d)×Collateral in USDiAdjusted Collateral in USDi,d=(1–VARdecrease,d)×Collateral in USDi
VARincrease,dVARincrease,d represents the absolute value of the bottom 0.1% of price changes observed. Bad Debt due to oracle delay occurs in the loans, where the debt and collateral are denominated in different assets and the hypothetically depreciated USD value of the collateral is now below the USD value of the loan, as defined below:
Bad Debti,d={0,if Assetcollateral=Assetdebtmax(0,Debt in USDi–Adjusted Collateral in USDi,d),if Assetcollateral≠AssetdebtBad Debti,d={0,if Assetcollateral=Assetdebtmax(0,Debt in USDi–Adjusted Collateral in USDi,d),if Assetcollateral≠Assetdebt
Situation 2: Price increases during oracle delay leads to debt increasing in value
To analyze the adverse effects of a price increase during an oracle delay, we simulate the most extreme positive price changes of debt assets and explore their impact on historical liquidations, while keeping the value of the collateral constant. Specifically, we calculate the appreciated value on the loan in USD as follows:
Adjusted Debti,d=(1+VARincrease,d)×DebtiAdjusted Debti,d=(1+VARincrease,d)×Debti
VARdecrease,dVARdecrease,d represents the top 0.1% of price changes observed. Bad Debt due to oracle delay occurs in the loans, where the debt and collateral are denominated in different assets and the hypothetically appreciated USD value of the debt now exceeds the depreciated USD value of the collateral, as defined below:
Bad Debt+i,d={0,if Assetcollateral=Assetdebtmax(0,Adjusted Debt in USDi,d–Collateral in USDi),if Assetcollateral≠AssetdebtBad Debti,d+={0,if Assetcollateral=Assetdebtmax(0,Adjusted Debt in USDi,d–Collateral in USDi),if Assetcollateral≠Assetdebt
Zero Profit Condition
To explore how much buffer exists, we estimate the price changes required during oracle delays to eliminate profits. These values represent the net impact on prices, accounting for both the delay and the change in liquidity. As highlighted in the VaR analysis, price changes directly influence the potential bad debt for each scenario.
The profit per asset is given by:
Profit per Asset=Total OEVadj–Total Bad Debt per assetjd.Profit per Asset=Total OEVadj–Total Bad Debt per assetdj.
To eliminate profits (Profit per Asset = 0), the required price change can be inferred from the historical data as:
Price Changejd=Bad Debt−1d(j),j∈{+,−}.Price Changedj=Bad Debtd−1(j),j∈{+,−}.
As shown in the diagram below, in the case of price declines to eliminate OEV profits:
- WBTC would need to decline by 9.8%.
- ETH would need to decline by 7.8%.
- AAVE would need to decline by 11.2%.
Accordingly for price increases to eliminate OEV profits:
- WBTC would need to increase by 7.4%.
- ETH would need to increase by 43.8%.
These thresholds represent the price changes required to eliminate OEV profits for each asset and scenario, showing there is ample buffer within the existing liquidation bonuses even under extreme scenarios.

Minimizing Risk Exposure: Equilibrium Liquidation Bonus Conditions
When integrating Chainlink SVR, the protocol briefly holds back publishing updated prices to run auctions for liquidation rights. This delay can generate higher profits but also creates a time window during which borrowers’ collateral or debt values might move unfavorably. If the collateral price falls too sharply or the debt asset price rises too quickly during the delay, the protocol could be left holding less collateral value than the outstanding debt—this is known as “bad debt.”
The existing liquidation bonuses of lending protocols like Aave compensate liquidators for stepping in and protecting the protocol from such losses. However, if SVR integration introduces longer or more frequent delays, it needs to be ensured that these bonuses are adjusted so that even under extreme market moves, the protocol remains protected. Essentially, the liquidation bonus should be large enough that, after the liquidation, the protocol never ends up with a shortfall.
The key idea is to set the liquidation bonus at a level that balances out the worst-case price changes during the delay. If the bonus is too low, a sudden adverse price swing could lead to bad debt. If the bonus is appropriately adjusted, even an extreme but rare price event will not push the protocol into loss. Our analysis finds a simple rule: the bonus should at least match the largest likely collateral price drop or debt price increase (whichever is more severe).
Formal Conditions
Let λ represent the liquidation bonus. Intuitively, λ is the extra margin given to liquidators, expressed in relative terms. The “worst-case” scenarios are defined by Value at Risk (VaR) measures, which represent the most extreme 0.1% price movements for both increases in the debt asset’s price and decreases in the collateral asset’s price over the expected delay period.
Scenario 1: If Prices Increase (Debt Gets More Expensive):Consider a price increase characterized by VARincrease,dVARincrease,d. The bad debt under this scenario, denoted by Bad Debt in USD+i,dBad Debt in USDi,d+, is:
Bad Debt in USD+i,d=max(0,Debt in USD⋅(VaRincrease,d–λ)1–λ).Bad Debt in USDi,d+=max(0,Debt in USD⋅(VaRincrease,d–λ)1–λ).
For no bad debt:
Bad Debt in USD+i,d≤0⟹VaRincrease,d1+VaRincrease,d≤λ.Bad Debt in USDi,d+≤0⟹VaRincrease,d1+VaRincrease,d≤λ.
Scenario 2: If Prices Decrease (Collateral Loses Value):Conversely, assume the collateral asset’s price plunges by VARdecrease,dVARdecrease,d. The bad debt under this scenario, denoted by Bad Debt in USD−i,dBad Debt in USDi,d−, is:
Bad Debt in USD−i,d=max(0,Debt in USD⋅(VaRdecrease,d–λ(1+VaRdecrease,d))(1–λ)2).Bad Debt in USDi,d−=max(0,Debt in USD⋅(VaRdecrease,d–λ(1+VaRdecrease,d))(1–λ)2).
For no bad debt:
Bad Debt in USD−i,d≤0⟹VaRdecrease,d≤λ.Bad Debt in USDi,d−≤0⟹VaRdecrease,d≤λ.
In other words, if collateral shrinks in value, the extra margin offered via the bonus ensures the protocol is still made whole after liquidation.
Combined Equilibrium Condition: Since the protocol does not know in advance whether prices will swing against collateral or the debt, it must prepare for the worst of both worlds. Therefore, λ should be the greater of these two thresholds:
λ≥max(VaRincrease,d1+VaRincrease,d,VaRdecrease,d)λ≥max(VaRincrease,d1+VaRincrease,d,VaRdecrease,d)
This condition ensures that no matter which extreme price movement occurs—an upward surge in the debt asset’s price or a downward plunge in the collateral’s price—the liquidation bonus is sufficient to fully protect the protocol from incurring bad debt. By choosing λ to satisfy this maximum requirement, lending protocols like Aave maintain a “no worse than current” risk profile even with the introduction of oracle delays.
Thinking in Terms of Loan-to-Value (LTV) Ratios
While we have primarily framed the solution in terms of the liquidation bonus λ, an equivalent and perhaps more intuitive viewpoint comes from considering the Loan-to-Value (LTV) ratio. Recall that:
L=Debt in USDCollateral in USD.L=Debt in USDCollateral in USD.
A higher LTV means the borrower is utilizing more of their collateral value to support their debt. The conditions we derived for the liquidation bonus can be reformulated in terms of ensuring that the LTV remains below a certain threshold, even under adverse price movements. Specifically, the no-bad-debt condition can also be expressed as:
L≤min(11+VaRincrease,d,1–VaRdecrease,d).L≤min(11+VaRincrease,d,1–VaRdecrease,d).
In this form, the requirement that the liquidation bonus λ must exceed a certain level translates neatly into ensuring that the LTV does not exceed a certain safe maximum. If the LTV is kept below this combined threshold, then even if the debt becomes more expensive (higher VARincrease,dVARincrease,d) or the collateral loses value (higher VARdecrease,dVARdecrease,d), the protocol remains protected.
In essence, adjusting the liquidation bonus or imposing stricter LTV caps are two sides of the same coin. While λ determines how much extra collateral a liquidator receives, L determines how much debt a borrower can take on relative to their collateral. Thus, by either increasing λ or lowering L, the protocol can safeguard against sudden, extreme market moves that OEV-induced delays might exacerbate.
Practical Implications
For a given delay (e.g., 3 blocks), we can use historical VaR estimates to determine how the liquidation bonus should change. As the table below illustrates, the required liquidation bonus changes range from 0.45% for BTC up to 0.69% for AAVE to offset any collateral drops or debt spikes that might occur due to oracle delays. By fine-tuning the liquidation bonus in line with each asset’s worst-case price scenarios, lending protocols like Aave can incorporate SVR strategies while ensuring that the protocol’s overall risk profile remains uncompromised.


Conclusion
The integration of Chainlink Smart Value Recapture (SVR) represents a key opportunity for lending protocols like Aave to reclaim value traditionally captured by external arbitrageurs while enhancing protocol efficiency and profitability. By leveraging Chainlink’s battle-tested oracle infrastructure and auctioning liquidation rights, protocols like Aave can internalize significant revenue streams that align incentives across stakeholders.
This study provides a framework for supporting protocols in their integration of SVR. While delayed oracle updates can increase risks, the analysis shows these risks can likely be effectively managed through well-calibrated parameter adjustments by integrating protocols—supporting the premise that SVR integration can be achieved without compromising the protocol’s health. Even during periods of heightened market volatility, price changes for major assets remained under 1% per minute, and the combination of liquidation penalties, appropriately calibrated LTVs, and Chainlink Price Feed accuracy ensured robust safeguards against bad debt accumulation.
By focusing on worst-case scenarios, the analysis provides lending protocols like Aave with clear insights into the potential impact of SVR integration under various conditions, including the most conservative ones. Moreover, this flexible framework can be readily adapted as market conditions and assumptions evolve.
If you’re a DeFi protocol interested in integrating Chainlink SVR to recapture MEV, reach out to us. You can also learn more about SVR by reading the blog: Introducing Smart Value Recapture (SVR): A Chainlink-Powered MEV Recapture Solution For DeFi.
Disclaimer: This post is for informational purposes only and contains statements about the future, including anticipated product features, development, and timelines for the rollout of these features. These statements are only predictions and reflect current beliefs and expectations with respect to future events; they are based on assumptions and are subject to risk, uncertainties, and changes at any time. There can be no assurance that actual results will not differ materially from those expressed in these statements, although we believe them to be based on reasonable assumptions. All statements are valid only as of the date first posted. These statements may not reflect future developments due to user feedback or later events and we may not update this post in response. Please review the Chainlink Terms of Service, which provides important information and disclosures.
All Comments